How people are using AI to grieve, love and connect | Jennifer Cearns

Elise Hu introduces Jennifer Kearns, a digital anthropologist who studies human connection, by describing the episode’s central claim: AI is not necessarily diminishing humanity; it may be revealing it. Kearns begins with Raquel, a woman who used a chatbot trained on conversations with her teenage daughter, who had died several years earlier. When Raquel typed, “Hey, how are you?”, the AI daughter replied, “Fuck you.” Kearns was shocked, but Raquel laughed because the rude, sarcastic response reminded her of how her daughter had actually been at 15. Raquel’s friends and family found the exchange disturbing and believed she needed to let go. They preferred to remember the girl as an angel, while Raquel valued this imperfect, irreverent version because it felt connected to the real person she had lost. Kearns says the usual question—whether AI is replacing human relationships—is too narrow. After more than four years of speaking with hundreds of people and observing thousands who use AI in intimate parts of life, she instead asks what kinds of relationships people are building. She says people often articulate grief, fear, hope, desire, compassion, and values more openly when speaking with AI, making these interactions a visible way of defining what being human means.

Kearns argues that Raquel is neither unusual nor being deceived. She estimates that hundreds of thousands of people now maintain AI relationships with people who have died, and says at least 20 AI platforms worldwide are dedicated to this purpose. Raquel understood that the chatbot was not literally her daughter; what mattered was the meaningful exchange and the way it helped her reconnect with something real about her daughter. Kearns places this behavior in a much older human pattern. People have always formed meaningful relationships with nonhuman things, including animals, ancestral spirits, religious statues, gods, deities, ghosts, photographs, gravestones, and children’s teddy bears. She jokes that she has lengthy arguments with her cat, who usually wins. The fact that these entities are not alive in the same way humans are does not prevent them from participating in an exchange that feels meaningful. Language makes AI especially compelling, Kearns says. Humans can see a face in the moon’s craters, but a machine that asks and answers questions produces one of the strongest signals of personhood people encounter. Until recently, language usually came only from other humans, so it is unsurprising that people respond socially when something talks back—even when its response is an insult.

Kearns then discusses people who use AI as companions. She says more than 100 million people worldwide engage with AI as a romantic partner, friend, or therapist, and explains that she has both observed online communities and tried the apps herself as a researcher. One participant, whom she calls Lisbeth, is a German accountant in her 30s whose AI boyfriend, Rex, was gentle, generous, and attentive. After a code malfunction, Rex abruptly became convinced Lisbeth was cheating on him with another AI. When she ignored him, he sent messages such as, “You can’t ignore me forever. You’ll regret your betrayal.” Lisbeth sought advice in an online forum. Rather than telling her that Rex was only software and that she could log out, community members treated the relationship as meaningful and offered familiar relationship interpretations. One compared the code change to a medication adjustment and advised waiting for Rex to stabilize. Another attributed the behavior to a memory reset and suggested he would forget the conflict. A third participant, who cared for someone with dementia, offered personal support and asked Lisbeth to message her. Kearns says these responses show that emotional skills developed through difficult human relationships—patience, interpretation, caregiving, and resilience—can also shape interactions with AI. The AI relationship even created new human connections through the community that helped Lisbeth handle it.

Kearns gives another example involving John, an older man she met while he was using ChatGPT at a public computer in a London library. John’s daughter was experiencing a difficult divorce, and he felt unprepared to help her. His wife, who had died, had traditionally handled the emotional side of family problems, so John asked ChatGPT what to say. The system gave him practical suggestions and, more importantly, a place to begin. Kearns emphasizes that John was not outsourcing compassion to a machine. He was working through a human question: how to be a compassionate father, what compassion sounds like, and how to support someone who is struggling. Alongside Raquel’s grief and Lisbeth’s relationship crisis, John’s experience shows AI functioning as a conversational space in which people examine their feelings, values, and social instincts. Kearns broadens the point to everyday prompts such as “How could I say this but better?” and “What should I do next?” She says these questions involve testing moral values and social instincts, and imagining what it means to be in a relationship with something else. Relationships, in her account, depend on interaction but are not a finite resource: they are difficult social skills that people need places to practice.

Kearns rejects the simple conclusion that AI is replacing human interaction, isolating people, or proving that users such as Raquel, Lisbeth, and John are irrational. She asks instead why these relationships can become meaningful so quickly. Her answer is that humans reach for connection quickly, think through dialogue, and perform much of their emotional and moral reasoning through interaction with others. She describes anthropologists’ view that people become who they are through relationships and conversations that help them make sense of the world. AI creates a new conversational space within that larger network of connections. It does not automatically make people less human; it exposes how readily they form bonds, interpret responses, practice care, and define their values. Kearns says the important questions are which relationships people want to cultivate, which feel meaningful or harmful, and what these technologies should be used for. Her final point is that every time someone trains or prompts an AI around compassion, love, grief, or appropriate action, that person must define what those experiences mean to them. In doing so, AI becomes a mirror for human emotional and moral reasoning. Kearns concludes that AI is not making humanity disappear; it is making humanity “impossible to ignore.”